Skip to main content
Messaging is the heart of your automation. In First Touch, you don’t write static templates - you build dynamic AI Prompts. Unlike a standard “mail merge” that just swaps a name, First Touch uses a Large Language Model (LLM) to generate unique, context-aware messages for every single prospect. This guide will teach you how to control that AI to speak in your voice.

​
The Messaging Editor

When you add an Email, Call, or LinkedIn Message step to a flow, you will see the Messaging Editor. This is your command center for outreach content.

​
Choosing Your Mode

You can toggle between two modes for generating content using the “Use First Touch messaging” switch:

AI Generated (Default)

Best for speed. When enabled, First Touch selects a battle-tested default prompt based on your specific Signal (e.g., “Job Posting”) or Audience. The AI automatically incorporates relevant context like funding news or hiring trends without you needing to write a single word.

Custom Template

Best for control. Disable the toggle to open the Rich Text Editor. Here, you define the specific structure, tone, context, and variables you want the AI to use. This is where “Prompt Engineering” happens.

​
Building Your Prompt (Custom Mode)

In Custom Mode, you aren’t writing the final email - you are writing the instructions for the AI. A strong prompt mixes three elements: Static Text, Data Variables, and Magic Wand Instructions.

​
1. Data Variables ({})

Variables inject raw data from your CRM or research directly into the prompt. This gives the AI the “facts” it needs to write the message.
  • How to add: Press the { key on your keyboard or click the icon in the toolbar.
  • Navigation: A menu will appear. Use your arrow keys to browse categories:
    • Prospect: Basic info like {prospect_first_name}, {prospect_role}, {prospect_location}.
    • Company: Details like {company_name}, {prospect_company_description}, {funding}.
    • Sender: Your own info (e.g., {sales_company_name}).
    • Flow: Advanced context like {web_research_answer} (more on this below).
    • HubSpot: Any mapped field from your connected CRM.
Pro Tip: If you used an AI Researcher node earlier in the flow, you can inject the specific answer it found using the {web_research_answer} variable. This allows you to reference highly specific data, like a company’s revenue model or recent challenges.

​
2. The “Magic Wand” Instructions ([])

This is the most powerful feature in the editor. Instead of just pasting data, you can tell the AI how to think or what to do with that data.
  • How to use: Click the Magic Wand icon in the toolbar or type [ to insert an instruction block.
  • Syntax: Inside the brackets, write a natural language command.
Examples of effective instructions:
  • [Write a casual 1-sentence icebreaker connecting their role as {prospect_role} to the challenges of scaling a team]
  • [Mention the recent news about {recent_news} and congratulate them]
  • [Transition to our value proposition: {sales_company_value_proposition}]
  • [Ask for interest in a 15-minute demo (Soft CTA)]

​
3. Static Text

You can type normal text anywhere in the editor. Use this for parts of the message that should remain exactly the same for everyone, such as:
  • Your sign-off (“Best regards,”)
  • Specific links to your calendar or website.
  • Legal disclaimers or unsubscribes.

​
Example Framework

Here is how a complete prompt might look in the editor:
Hi {prospect_first_name}, [Write a short observation about their company {company_name} based on their industry {prospect_company_industry}] I saw that you recently {recent_news}. [Explain how our product helps companies in this situation] [Ask if they are open to a conversation] Best, {sender_first_name}

​
Preview & Testing

You should never publish a flow blindly. The Preview tab allows you to see exactly how your prompt translates into a real email.
  1. Open Preview: Click the Preview tab at the top of the editor.
  2. Select a Prospect: Use the dropdown to select a sample prospect (real or dummy data).
  3. Generate: First Touch will run your prompt against that prospect’s data in real-time.
  4. Iterate: If the AI sounds too formal, or missed a variable, switch back to Edit, adjust your instructions (e.g., add [Keep it under 50 words]), and preview again.

​
LinkedIn Messaging Strategy

LinkedIn requires a different approach due to platform etiquette and technical limits.

Connection Requests

Keep it brief.
  • Recommendation: We strongly suggest sending connection requests without a note. Acceptance rates are often higher for blank requests as they feel less “salesy.”
  • Character Limit: If you do add a note, it must be under 300 characters. First Touch will automatically trim messages that exceed this limit, potentially cutting off your sentence.

Follow-up Messages

The “Accepted” Trigger.
  • A LinkedIn Message step in a flow will only execute if the prospect accepts your connection request.
  • If the prospect is already a 1st-degree connection, the message sends immediately.
  • If the request is pending, the flow waits. You can chain up to 5 follow-up messages to be sent after acceptance.
Safety Limits: To protect your account, we recommend setting daily limits in Settings > LinkedIn to 20-40 actions per day.

​
Advanced Strategies

Don’t guess - know.Add an AI Researcher node before your email step. Configure it to answer a specific question, such as: “What is this company’s pricing model?” or “Who are their main competitors?”Then, in your email prompt, inject the {web_research_answer} variable.
  • Prompt: [Mention their competitor {web_research_answer} and ask how they differentiate]
  • Result: A message that proves you did your homework.
Hyper-segmentation.Create different message paths for different personas. Use a Conditional Split card to check a variable (e.g., Job Title CONTAINS “Marketing”).
  • Yes Path: Route to a marketing-focused message.
  • No Path: Route to a general leadership message.
Quality Control.For high-value prospects (e.g., Fortune 500 CEOs), enable “Human approval required” on the email step settings.This pauses the flow right before sending and creates a Task for you. You can review the AI-generated draft, manually tweak it if necessary, and hit “Approve” to send.